|Publication number||US7065576 B2|
|Application number||US 09/965,691|
|Publication date||Jun 20, 2006|
|Filing date||Sep 27, 2001|
|Priority date||Sep 27, 2001|
|Also published as||CN1409498A, CN100448180C, DE60209448D1, DE60209448T2, EP1298837A2, EP1298837A3, EP1298837B1, US20030060216|
|Publication number||09965691, 965691, US 7065576 B2, US 7065576B2, US-B2-7065576, US7065576 B2, US7065576B2|
|Inventors||Ibrahim Mostafa Kamel, Li Zou|
|Original Assignee||Matsushita Electric Industrial Co., Ltd.|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (18), Non-Patent Citations (3), Referenced by (32), Classifications (21), Legal Events (3)|
|External Links: USPTO, USPTO Assignment, Espacenet|
The present invention relates generally to distributed information systems. More particularly, the invention relates to a scalable multicast algorithm for mobile agents. Applications for the invention include the intelligent transportation system where smart vehicles are connected to a network and periodically send state information, such as position, speed, traffic and weather conditions, and the like. The invention has further utility in the intelligent cooperative transportation system, in which vehicles cooperate to deliver merchandise to its destination. The principles of the invention can also be extended to other mobile object applications, including multi-user network games.
Distributed information systems are becoming increasingly popular as more and more devices establish connectivity on the internet. One major type of distributed information system involves clients that are only interested in information of its immediate geographical vicinity. The technique of the invention aims to efficiently discover and retrieve such information. One particularly promising application involves the intelligent transportation system. Once the majority of all vehicles are equipped with wireless internet connectivity capability, software applications running on the vehicular computer systems can be used to establish a distributed information system.
In such a system each vehicle would serve as both a data source and a data destination, gathering information about the location of each vehicle, its speed, direction of travel, and other state information such as traffic and weather conditions. In effect, each vehicle would gather information about its state and would broadcast that information to interested vehicles, which would make use of the information.
One common way to construct a distributed information system for this kind of application is to establish a central server which collects information from all users (e.g., all vehicles). When a user wants to get information about conditions in a particular region, the user sends a request to the server and the server then transmits the results back to the requesting user. This approach has scaling problems, as the number of users in the application increases, the network bandwidth and processing resources required by the central server increase rapidly. The central server soon becomes a bottleneck, and a single point of failure for the system.
Another common technique is to have all users join a common multicast group through which they broadcast all information. After receiving information from the multicast group, each user must filter out any messages that it considers superfluous. In the traffic application, for example, a user on the upper east side of the city might have no interest in traffic conditions on the lower west side and might therefore configure the client application software to filter out lower west side information as superfluous.
While the common multicast approach solves the single point of failure problem associated with a single server, it does so at a significant cost of increasing network bandwidth requirements and also increasing the processing resources consumed by each user's onboard computer.
The present invention addresses the problem by providing multiple multicast groups. The system divides an area into regions or cells and associates each region or cell with a multicast group address. In this way, users that are only interested in information in their immediately surrounding area, or vision domain, discover information of interest through cooperation with other nearby users. Thus a user on the upper west side would acquire information from other users in that region or cell, using the multicast group address associated with that cell. Meanwhile, information concerning conditions in the lower west side cell would be communicated using a different multicast group address. Thus the users on the upper east side would not need to filter out lower west side information. Of course, if a particular user is interested in obtaining information about another region or cell, the well-designed system should be prepared to deliver it. The present invention makes this possible by using a dynamically reconfigurable quad-tree data structure that stores the current configuration and relationship of all cells or regions. While the quad-tree data structure is presently preferred, other data structures such as K-d trees or the like may be used. The data structure is maintained and updated by a partitioning server, or servers, which can be a central server or distributed servers, that determines the optimal size and grouping of cells based on the ever-changing needs of the users.
As the vehicles for mobile objects move from place to place, the partitioning server acquires information about this movement and reconfigures the quad-tree data structure to optimize performance. Thus regions with a high-density of users will be broken into a larger number of cells by partitioning one or more existing cells into smaller ones. Conversely, regions that are sparsely populated may be assigned to larger cells by merging smaller contiguous cells that are less populated.
According to one aspect of the invention, the dynamic quad-tree partitioning algorithm assigns a coordinator for each cell that is elected by the users of that cell. The coordinator then assesses the cost associated with splitting or merging its cell based on local knowledge of usage conditions. By having one elected coordinator perform this function, the other users within the cell do not have to expend computational resources gathering this information. The coordinator sends the cost information to the partitioning server. The partitioning server uses an algorithm to determine which cells should be split and which cells should be merged to minimize the global cost, thereby optimizing the performance of the system. The partitioning server then broadcasts the new partition scheme to all users. In this way, the users are all advised of which cells are available to communicate with. Thus, if a user on the upper west side wishes to obtain information from the lower east side, that user can identify which cell is currently associated with the lower east side that may contain information of interest.
The invention thus provides a hybrid solution that uses both distributed components (multicasting, coordinator) and centralized components (partition server) to dynamically partition an area so that the mobile objects are grouped efficiently. For a more complete understanding of the invention, its objects and advantages, refer to the following description and to the accompanying drawings.
The present invention will become more fully understood from the detailed description and the accompanying drawings, wherein:
The dynamic multicast grouping algorithm of the present invention has many uses. To teach the principles of the invention, a distributed traffic condition discovery application is illustrated in
Two mobile entities 14 and 16 have been illustrated in greater detail in
Collectively all of the communicating entities compile a body of global information 22 that is made up of all of the collective bodies of local information, such as local information 24 associated with entity 16. In the presently preferred embodiment the global information 22 is maintained in a distributed fashion. Thus it is not necessary to upload and store an entire copy of the global information in a central database. Rather, the global information exists in a distributed state, in the form of a collection of local information bodies. Each mobile entity can access the information that is relevant to that mobile entity, according to its specific vision domain. Because the local entities control their respective vision domains, any mobile entity can, if desired, acquire information about any region within the geographic area. In a traffic application, typically the mobile entity will be interested in traffic conditions nearby.
Mobile entities communicate with other mobile entities to obtain information, typically based on geographic considerations. Entities communicate with other entities that are physically present within a geographic region of interest, or that have acquired information about a geographic region of interest. The mobile entities determine which other entities they wish to communicate with using a common data structure that reflects how the mobile entities are geographically grouped or arranged. To give all mobile entities a common basis for organizing how they are grouped, the presently preferred embodiment uses a quad-tree data structure that is dynamically updated to reflect changing distribution patterns of the entities as they move about. In the preferred embodiment, the quad-tree data structure is generated centrally at the partition server, based on information collected and evaluated locally. Each time the quad-tree is revised, a copy of the quad-tree data structure is shared by with entities.
Each node or cell within the quad-tree data structure has an assigned multicast group address. A mobile entity wishing to obtain information about conditions associated with that cell simply joins the multicast group. The quad-tree data structure thus guides the mobile entities in determining which multicast groups to join, based on that entity's current vision domain.
Each mobile entity acts as both a data source and a data destination in the preferred embodiment. This has been illustrated diagrammatically in
In the presently preferred embodiment each mobile entity has a computer system running a client application that is functionally described in
So that the client application can also provide information to the body of global information 22, the client application includes a data source port 42. Local information is stored in the local data store 44. As illustrated, this local data store supplies information to the data source port 42. It also receives information from the data destination port 40 after first being processed by an information filter 46. The information filter screens out superfluous information, such as information that does not pertain to the vision domain of the mobile entity.
The local data store is preferably stored in computer readable memory that is administered by the client application. The local data store preserves information about the mobile entity and its membership within a home cell of the quad-tree, as represented diagrammatically at 50. The local data store also preserves other local state information that is application specific. As shown diagrammatically at 52, the local state information may include data on the location of the mobile entity, its speed and other relevant positional data, and other conditions such as temperature, wind speed and other environmental conditions.
In the presently preferred embodiment, each mobile entity has the possibility of serving as the coordinator for the cell in which that entity is located. It is the coordinator's responsibility to assess congestion conditions within that cell, to assess whether that cell should be subdivided or merged with other cells to improve performance. In the preferred embodiment a single coordinator is elected for each cell, based on where a mobile entity is located physically within the cell. Other criteria may be used instead, if desired. The client application employs a processing module 54 that accesses information from local data store 44 regarding the current quad-tree configuration and the mobile entity's location within a particular cell to determine whether that entity meets the qualifications to be the coordinator. All other mobile entities have the same module 54 and thus are also able to determine whether they qualify as the coordinator. The coordinator election rules are designed so that one and only one coordinator is elected per cell in the presently preferred embodiment.
Because each mobile entity is potentially capable of meeting the requirements to be elected coordinator for a particular cell, the processing module 54 includes a processing routine to evaluate cost information reflecting whether the cells should be subdivided or merged. This cost information is evaluated at the cell level and is then communicated to the partitioning server, as will be more fully discussed below. The presently preferred information assesses cost based on the number of multicast group joins that have occurred within a given cell and the number of superfluous messages the members of that cell have had to filter out or reject. The join cost is incurred by a user joining the multicast groups associated within cells. The superfluous message cost is incurred by counting the number of unwanted messages received from a multicast group. Subdividing a cell into a smaller number of child cells establishes a finer grid that gives the system more capability of discerning among users and thus reduces the number of superfluous messages. On the other hand, subdividing cells into increasingly smaller child cells increases the need for users to join additional multicast groups in order to fulfill their vision domain. Thus, there is a tradeoff between the number of multicast group joins and the number of superfluous messages. This tradeoff is directly related to the density of users within a given area.
As noted above, the presently preferred embodiment uses a quad-cell data structure to represent an area, and represent how that area is subdivided into regions or cells. As illustrated in
According to the dynamic repartition algorithm implemented by the invention, the size and configuration of the quad-tree cells will vary dynamically as mobile entities move about. The basic partition scheme by which the quad-tree is configured at any given time is calculated by a partitioning server, using information passed to it by the elected coordinators of each cell.
The partitioning server uses a predefined algorithm to determine a new partition scheme, using information received from all coordinators of the various cells represented in the quad-tree. It broadcasts a new partition scheme 82 to the mobile users, either by broadcasting to all user multicast groups, or by relaying the information through a single user, such as coordinator 74. The system may have a dedicated coordinator multicast group for this purpose. The new partition scheme represents a new quad-tree that is stored by each mobile user (in local data store 44,
The presently preferred partitioning server uses a greedy algorithm 84 to generate the new partition scheme by modifying the existing quad-tree data structure based on locally estimated cost data. To generate the initial quad-tree data structure a top-down greedy algorithm is presently preferred. A flow chart of this greedy algorithm is illustrated in
To understand the coordinator election process, first refer to the large cell 200. One of its child cells 202 is shown subdivided into grandchild cells 204, 206, 208 and 210. Each of the grandchild cells 204–210 will elect one coordinator. The winning candidate is the one that is closest to the center of that child's parent cell. In this case, the center of parent cell 204 has been marked with an X. Because grandchild cell 204 contains only one entity, that entity is, by default, closest to the center X of the parent cell. Mobile entity 212 of grandchild cell 206 is closer to the center X than any other entity in that cell. Thus entity 212 is elected as coordinator for cell 206. Similarly, entities 214 and 216 win the closest to the center contest in their respective cells and are thus elected as coordinators.
While the closest to center technique serves to determine one coordinator for each cell, other techniques for electing the coordinator may be employed in the alternative. Because the system is designed to expect mobile entities to move about, the election of coordinator process is repeated at a sufficient frequency to capture the dynamic nature of the distributed information system.
Referring back to
As noted above, the presently preferred embodiment uses a top-down greedy algorithm to generate the initial partition scheme for the quad-tree. The greedy algorithm is illustrated in
An example of a bottom-up greedy algorithm has the following steps:
In use, the mobile entities use their respective vision domains and the current quad-tree to effect a state disseminating protocol. The protocol is designed to ensure that the state message originating from each mobile entity reaches all other interested entities. The protocol includes the following steps:
Dynamic repartitioning is effected in the presently preferred embodiment by starting with the static partitioning generated using the greedy algorithm of
While the above algorithm does rely on a partitioning server to decide the new partition scheme, the server carries very little computational burden because the cost estimation is performed distributively by the coordinators.
The dynamic partitioning algorithm of the invention was tested using computer simulation with real geographical data and was found to reduce message costs by as much as 80 percent. In the simulation, the performance of the dynamic quad-tree partition algorithm was tested and compared with cell-based protocols and static quad-tree partitions. The square simpulation area is assumed to have unit side lengths. In the simulation, it was assumed that all users had the same circular vision domain. The radius of that vision domain was set at 0.01. All users randomly chose a direction and moved a constant step toward that direction. If the movement would take the user out of the area, the user was “bounced back” into the area such that the number of users remained constant. In the simulation, we assumed that all users would have the same step size. The step sizes chosen were 0.001, 0.002, 0.005 and 0.01, respectively. To reduce the amount of data, we sampled one data point every 100 steps. A new partition was created when the new partition would decrease the message cost by 20 percent. The results demonstrated that the dynamic quad-tree partition algorithm would save as much as 80 percent of the message cost.
While the invention has been described in its presently preferred implementation, it should be understood that the invention is capable of modification and is applicable to a variety of different uses, all within the spirit of the invention as set forth in the appended claims.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US5457680 *||May 18, 1993||Oct 10, 1995||International Business Machines Corporation||Data gateway for mobile data radio terminals in a data communication network|
|US5848373 *||Jul 18, 1997||Dec 8, 1998||Delorme Publishing Company||Computer aided map location system|
|US6032051 *||Dec 1, 1997||Feb 29, 2000||Telefonaktiebolaget L/M Ericsson||Wireless mobile comunication devices for group use|
|US6128648 *||Nov 23, 1994||Oct 3, 2000||International Business Machines Corporation||Information handling system and method for maintaining coherency between network servers and mobile terminals|
|US6336134 *||Dec 30, 1999||Jan 1, 2002||International Business Machines Corporation||Dynamic clients, dynamic partitions, locking, and migration capability for distributed server for real-time collaboration|
|US6424840 *||Nov 5, 1999||Jul 23, 2002||Signalsoft Corp.||Method and system for dynamic location-based zone assignment for a wireless communication network|
|US6477149 *||Jan 27, 1999||Nov 5, 2002||Nec Corporation||Network system and method of controlling multicast group participation of mobile host|
|US6574633 *||Feb 10, 2000||Jun 3, 2003||Honeywell International Inc.||Method for dynamically grouping limited range physical entities in a topological space|
|US6754463 *||Nov 20, 2002||Jun 22, 2004||Ricoh Company, Ltd.||Image reading apparatus and method of accelerating drive of reading portion thereof|
|US6757262 *||Sep 15, 2000||Jun 29, 2004||Motorola, Inc.||Service framework supporting remote service discovery and connection|
|US6757433 *||Aug 28, 2000||Jun 29, 2004||Samsung Electronics Co., Ltd.||Object-based quad-tree mesh motion compensation method using greedy algorithm|
|US6847633 *||Feb 26, 1999||Jan 25, 2005||The Directv Group, Inc.||Internet-augmented radio port controller unit (RPCU) of personal acces communications systems (PACS)|
|US6917985 *||Mar 9, 2001||Jul 12, 2005||The Regents Of The University Of California||Core assisted mesh protocol for multicast routing in ad-hoc Networks|
|US20020026525 *||Apr 3, 2001||Feb 28, 2002||Armitage Grenville J.||Supporting mobile hosts on an internet protocol network|
|US20020058495 *||Dec 30, 1998||May 16, 2002||Albert Chow||Method and apparatus for providing neighborhood cordless services|
|US20020156578 *||Jun 10, 2002||Oct 24, 2002||Takeshi Kondou||Method and system for providing information for a mobile terminal and a mobile terminal|
|US20050213545 *||Feb 4, 2002||Sep 29, 2005||Choyi Vinod K||Micro-mobility network routing system and method|
|WO2001065885A1||Feb 9, 2001||Sep 7, 2001||Telefonaktiebolaget Lm Ericsson (Publ)||Reducing signaling traffic with multicasting in a wireless communication network|
|1||J. Habetha et al., "Outline of a Centralised Multihop ad hoc Wireles Network", Computer Networks, Elsevier Science Publishers B.V., Amsterdam, NL, vol. 37, No. 1, Sep. 2001, pp. 63-71.|
|2||M.A. Dru and S. Saada, "Location-Based Mobile Services: the Essentials" Alcatel Telecommunications Review, Jan. 1, 2001, pp. 71-76.|
|3||Mario Gerla et al., "Tree Multicast Strategies inMonile, Multihop Wireless Networks" Mobile Networks and Applications, vol. 4, Oct. 1999, pp. 193-207.|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US7532622 *||Jun 16, 2003||May 12, 2009||National University Of Singapore||Methods, devices and software for merging multicast groups in a packet switched network|
|US7937370||Feb 21, 2007||May 3, 2011||Axeda Corporation||Retrieving data from a server|
|US7966418||Feb 20, 2004||Jun 21, 2011||Axeda Corporation||Establishing a virtual tunnel between two computer programs|
|US8040840 *||Apr 30, 2008||Oct 18, 2011||Samsung Electronics Co., Ltd.||Method of merging cells in cognitive radio|
|US8055758||Aug 14, 2006||Nov 8, 2011||Axeda Corporation||Reporting the state of an apparatus to a remote computer|
|US8060886||Feb 12, 2007||Nov 15, 2011||Axeda Corporation||XML scripting of SOAP commands|
|US8065397||Dec 26, 2006||Nov 22, 2011||Axeda Acquisition Corporation||Managing configurations of distributed devices|
|US8108543||Apr 17, 2002||Jan 31, 2012||Axeda Corporation||Retrieving data from a server|
|US8291039||May 11, 2011||Oct 16, 2012||Axeda Corporation||Establishing a virtual tunnel between two computer programs|
|US8370479||Oct 3, 2006||Feb 5, 2013||Axeda Acquisition Corporation||System and method for dynamically grouping devices based on present device conditions|
|US8406119||Sep 29, 2006||Mar 26, 2013||Axeda Acquisition Corporation||Adaptive device-initiated polling|
|US8412206 *||May 21, 2010||Apr 2, 2013||Samsung Electronics Co., Ltd.||Communication system having network access structure|
|US8478861||Jul 6, 2007||Jul 2, 2013||Axeda Acquisition Corp.||Managing distributed devices with limited connectivity|
|US8752074||Oct 4, 2011||Jun 10, 2014||Axeda Corporation||Scripting of soap commands|
|US8762497||Dec 19, 2011||Jun 24, 2014||Axeda Corporation||Retrieving data from a server|
|US8769095||Dec 26, 2012||Jul 1, 2014||Axeda Acquisition Corp.||System and method for dynamically grouping devices based on present device conditions|
|US8788632||Oct 4, 2011||Jul 22, 2014||Axeda Acquisition Corp.||Managing configurations of distributed devices|
|US8806426||Jun 4, 2008||Aug 12, 2014||Microsoft Corporation||Configurable partitioning of parallel data for parallel processing|
|US8898294||Oct 3, 2011||Nov 25, 2014||Axeda Corporation||Reporting the state of an apparatus to a remote computer|
|US9002980||Sep 13, 2012||Apr 7, 2015||Axeda Corporation||Establishing a virtual tunnel between two computer programs|
|US9170902||Feb 20, 2013||Oct 27, 2015||Ptc Inc.||Adaptive device-initiated polling|
|US9491049||Jul 18, 2014||Nov 8, 2016||Ptc Inc.||Managing configurations of distributed devices|
|US9491071||Jun 27, 2014||Nov 8, 2016||Ptc Inc.||System and method for dynamically grouping devices based on present device conditions|
|US9591065||Jun 6, 2014||Mar 7, 2017||Ptc Inc.||Scripting of SOAP commands|
|US9674067||Oct 23, 2015||Jun 6, 2017||PTC, Inc.||Adaptive device-initiated polling|
|US9712385||Apr 6, 2016||Jul 18, 2017||PTC, Inc.||Managing configurations of distributed devices|
|US20030135594 *||Dec 6, 2001||Jul 17, 2003||Lin Xu||System and method for efficient distribution of multicastable services|
|US20040252690 *||Jun 16, 2003||Dec 16, 2004||Pung Hung Keng||Methods, devices and software for merging multicast groups in a packet switched network|
|US20060009207 *||Jun 30, 2005||Jan 12, 2006||Groundhog Technologies Inc.||Partitioning method for mobile communication network|
|US20090017828 *||Apr 30, 2008||Jan 15, 2009||Samsung Electronics Co., Ltd.||Method of merging cells in cognitive radio|
|US20090319992 *||Jun 4, 2008||Dec 24, 2009||Microsoft Corporation||Configurable partitioning for parallel data|
|US20110096736 *||May 21, 2010||Apr 28, 2011||Samsung Electronics Co., Ltd.||Communication system having network access structure|
|U.S. Classification||709/227, 455/519, 709/223, 709/228|
|International Classification||H04L12/56, G08G1/09, G08G1/13, H04B7/26, G06F15/173, H04L12/18, G06F15/16, H04W8/14, H04W4/08, H04W8/08|
|Cooperative Classification||H04L12/185, H04W8/14, H04W8/08, H04W4/08, H04L12/189|
|European Classification||H04L12/18M, H04W4/08|
|Sep 27, 2001||AS||Assignment|
Owner name: MATSUSHITA ELECTRIC INDUSTRIAL CO., LTD., JAPAN
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:KAMEL, IBRAHIM MOSTAFA;ZOU, LI;REEL/FRAME:012218/0414;SIGNING DATES FROM 20010913 TO 20010920
|Nov 18, 2009||FPAY||Fee payment|
Year of fee payment: 4
|Nov 20, 2013||FPAY||Fee payment|
Year of fee payment: 8